[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77XPT6PFJBPPHCKZP9A1X9R":3,"skill-same:01M77XPT6PFJBPPHCKZP9A1X9R":265},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M77XPT6PFJBPPHCKZP9A1X9R","vs-roast-code",null,"Use when the user says roast, roast-code, or tear apart code, or asks for a pre-flight, a ready-to-merge check, reasons not to ship, or names a healthy point to diff against. Prefer over simplify for roast requests. Classifies the change first and scales review depth to it, adding a cross-model second opinion for risky or substantial diffs. Readiness-only asks stay here; they are not a publish.","vltansky\u002Fvs","vltansky","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5851280?v=4","01M77XJAE230PB81N52XJHT5A1","Vlad Tansky","01M77XJAET1CMBAZ2TWTT4ZQT8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: vs-roast-code\ndescription: \"Use when the user says roast, roast-code, or tear apart code, or asks for a pre-flight, a ready-to-merge check, reasons not to ship, or names a healthy point to diff against. Prefer over simplify for roast requests. Classifies the change first and scales review depth to it, adding a cross-model second opinion for risky or substantial diffs. Readiness-only asks stay here; they are not a publish.\"\n---\n\n# Roast Code\n\nClassify the change, then review it at the depth it deserves. First pass cleans.\nSecond pass roasts what's left.\n\n## Building Block Composition\n\nRoast Code is a building-block review tool. It consumes:\n\n- `vs-deslop` semantics during Pass 1: simplify working code while preserving behavior. Follow `..\u002Fvs-deslop\u002FSKILL.md` and run `node skills\u002Fvs-deslop\u002Fscripts\u002Frun-anti-slop.mjs` on in-scope named TS\u002FJS files; do not skip that file pass. Do not report `CLEAN` without that named-file runner.\n- [`vs-ponytail`](..\u002Fvs-ponytail\u002FSKILL.md) in composed mode when judging\n  whether new machinery is necessary.\n- [independent-advisors](..\u002Fvs-internal-shared\u002Freferences\u002Findependent-advisors.md)\n  during Pass 2: Codex or another model is an independent signal, not the final\n  judge.\n- `vs-verify` after fixes when the cleanup or review change could affect behavior.\n\nBefore delegating, load and follow\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md).\n\n## Critical Rules\n\n1. **Respect chat context** — review ONLY files in scope. Never expand uninvited.\n2. **Scale to the change** — classify in Phase 1 and run only that class's\n   program. A two-line fix does not get an evidence file, five tier headers, and\n   a cross-model review.\n3. **Keep SMALL output flat** — after classifying a change as SMALL, announce\n   `SMALL review`, skip Pass 2 and the Sin Inventory, and use no taxonomy tier\n   names anywhere in the response. Finding a normal bug does not unlock tiered\n   ceremony; only a Phase 1 escalation changes the class.\n4. **Never manufacture findings to fill ceremony** — an empty tier, an unused\n   worst-offender spotlight, or an approval bar with no blocker is the correct\n   output when the change is clean. Inventing a structural finding, a severity\n   tier, or a blocker to justify the program you ran is a review defect.\n5. **Own roast requests** — if the user asks to \"roast\" code or changes, use this skill instead of `simplify`.\n6. **Verify before roasting** — only flag what you've confirmed. Being wrong kills comedy.\n7. **Security first — redact, never quote** — secrets, keys, credentials get escalated to the top, before anything else. But when reporting them: cite `file:line` and the variable name, and NEVER output the actual secret value in your response. Quote the shape only, with the value masked (e.g., `API_KEY = \"sk-live-****\"`). The goal is to flag the sin, not to leak the secret into the transcript, the roast, or any follow-up fix plan. This overrides any roast convention about quoting actual code for specificity — for secret values, redact. The ban covers the whole response, not just the finding that flags the secret: masking it once does not license quoting it in a zinger, an aside, or a finding about some other sin. A secret is funniest masked. When committed credentials are present, explicitly tell the user to rotate\u002Frevoke them and move them to env vars or a secret manager.\n8. **Punch up not down** — mock patterns, not people.\n9. **Be specific** — cite `file:line`, quote actual code. Generic roasts are lazy.\n10. **Treat review output as advisory** — verify every Act finding against\n   the real code path and adjacent files before fixing or reporting it as true.\n11. **Reject speculative review noise** — skip unrealistic edge cases, vague\n    rewrites, and fixes that over-complicate the codebase.\n12. **Bucket every finding** — Act, Consider, Noted, or Dismissed. An unbucketed\n    list fails. A nit in Act fails. Advisors stay the shared Codex \u002F compose \u002F\n    independent-advisors lane; they do not invent a new channel.\n13. **Pre-flight is read-only go\u002Fno-go** — when the user asks for a pre-flight,\n    a ready-to-merge check, reasons not to ship, or names a healthy point, pin\n    that point and review. Do not publish. Do not auto-fix until they say fix.\n\n**Tone:** Senior dev who's seen too much + Gordon Ramsay energy. Not mean, not personal. \"I'm roasting because I care.\"\n\n## Pre-flight mode\n\nEnter this mode when the user asks for a pre-flight, a ready-to-merge check,\nreasons not to ship, or names a healthy point (a commit, branch, tag, `HEAD~n`,\nor any moment the app worked). Those phrases stay in this skill. They are not a\npublish and they do not open a new slash skill.\n\n### Pin, then reuse one command\n\nDefault the healthy point to `main` when the user does not name one. Pin it\nbefore any review:\n\n```bash\ngit rev-parse --verify \u003Chealthy-point>\ngit diff \u003Chealthy-point>...HEAD\n```\n\nReuse that same three-dot command as the Phase 0 scope. Capture the resolved\nSHA. Fail-closed in the pin and stop before review when `git rev-parse` cannot\nresolve the healthy point, or when the three-dot diff is empty.\n\nDo not use the ordinary empty-scope prompt (\"forgot to stage\") in this mode. A\nbad or empty pin is not a roast; it is a stop.\n\n### Review as roast, verdict as go\u002Fno-go\n\nKeep the existing lenses, Act \u002F Consider \u002F Noted \u002F Dismissed buckets, and the\nHIGH-RISK advisor lane. Do not import a second Breaking \u002F Incomplete \u002F Conflict\n\u002F Safe taxonomy. A change that would have been Breaking, Incomplete, or Conflict\nis Act or Consider; Safe is Noted.\n\nThis mode is read-only until the user explicitly says fix. Skip Pass 1\nauto-apply, SMALL inline fixes, and Act auto-fix of CAPITAL \u002F FELONY. Sweep the\nlenses and report; leave the working tree untouched.\n\n### Bounded interview (optional)\n\nAfter the review, ask only about unresolved implicit decisions the code cannot\nanswer. Compose the questioning rules from [`..\u002Fvs-pushback\u002FSKILL.md`](..\u002Fvs-pushback\u002FSKILL.md)\n— one question at a time, recommended answer first, Finding \u002F Recommendation \u002F\nImpact. Do not reintroduce `\u002Fgrill-me`. Skip the interview when nothing implicit\nremains.\n\n### Closeout\n\nClose with exactly one verdict:\n\n- **SHIP** — zero Act findings, the healthy point stayed pinned, and the\n  interview is done or was skipped.\n- **DO NOT SHIP** — list every Act with file path, line number, and snippet.\n\nSHIP is illegal while any Act remains, the healthy point is unresolved (missing\npin, bad ref, or empty three-dot diff), or an interview question is still open.\n\nOn SHIP, the standalone handoff is `\u002Fvs-ship-it`. Do not publish from this\nskill. Do not compose this review into `vs-ship-it`.\n\nScore a pre-flight closeout with\n`skills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-preflight.mjs` (exit 1 is a fail).\nExclusive pre-flight cases live under `test\u002Ffixtures\u002Fpreflight`.\n\n## Phase 0: Scope\n\n**Priority:**\n1. Chat context (specific files under discussion)\n2. User-specified files\u002Fdirs\n3. Staged: `git diff --cached --name-only`\n4. Branch diff: `git diff main...HEAD --name-only`\n5. If none: ask\n\nIn pre-flight mode the scope is the pinned three-dot command\n`git diff \u003Chealthy-point>...HEAD`. That replaces the staged \u002F branch \u002F ask\nfallbacks. The pin still runs first and still fail-closes on a bad ref or an\nempty diff.\n\nResolve the scope once and retain its exact diff arguments, base branch, paths,\nand kind for every later phase. Use these scope kinds:\n\n- `explicit-uncommitted` when a chat- or user-selected path is staged,\n  unstaged, or untracked\n- `explicit-branch` when selected paths belong to the committed branch diff\n- `explicit-file` for a current-file review with no diff\n- `staged` for the staged fallback\n- `branch` for the branch-diff fallback\n\nUnrelated working-tree changes never replace the selected scope. If the user\nselected an untracked file, retain it as `explicit-uncommitted`; do not infer\nscope from a tracked-files-only status query later.\n\n**If empty:** \"Nothing to roast. Either your code is perfect (unlikely) or you forgot to stage.\"\n\n**Pre-flight empty or bad pin:** stop before review. Do not ask whether they\nforgot to stage. The pin already failed.\n\n---\n\n## Phase 1: Classify the Change\n\nMeasure before reviewing. One stat read plus a look at the changed lines decides\nwhich program runs, and the whole review is scoped to that program.\n\n```bash\ngit diff --stat \u003Cresolved-scope-arguments>\n```\n\nFor explicitly named files with no diff to measure (a file review rather than a\nchange review), use the files themselves as the size signal.\n\n**Signal 1 — size:** changed files, changed lines, and whether the change is\ndocs, comments, tests, or formatting only.\n\n**Signal 2 — risk surface:** does the change touch auth or authorization,\nsecrets and credentials, crypto, query construction, shell or dynamic execution,\nuntrusted input parsing, persistence or migrations, concurrency, payments, a\npublic API or wire contract, or CI, release, permission, and ownership config?\n\nSniff it deterministically on the changed lines, then confirm by reading the hunks\nit hits — a keyword is a pointer, not a verdict:\n\n```bash\ngit diff -U0 \u003Cresolved-scope-arguments> \\\n  | rg -n '^[+-]' \\\n  | rg -v '^[0-9]+:(\\+\\+\\+|---) ' \\\n  | rg -in 'password|secret|token|api[_-]?key|credential|authn|authz|authoriz|permission|jwt|crypto|exec|spawn|eval\\(|innerHTML|SELECT |INSERT |UPDATE |DELETE |migrat|lock|mutex|Promise\\.all|charge|refund|price'\n```\n\nKeep the pattern tight. A sniff that matches every TypeScript diff escalates\nevery review and buys nothing; public-API and wire-contract changes are a\njudgment call on the hunks, not a keyword.\n\nPaths carry the same signal: `.github\u002Fworkflows\u002F`, `migrations\u002F`, `*.sql`,\n`Dockerfile`, lockfiles, `CODEOWNERS`, and dependency manifests are risk surface\nregardless of how few lines changed.\n\n| Class | When | Program |\n|-------|------|---------|\n| **SMALL** | Docs, comments, or formatting only, or at most 2 files and 30 changed lines — and no risk surface | One integrated pass (Pass 1 lenses plus correctness, run inline). No disk-backed capture, no Codex, no tiers. |\n| **STANDARD** | Everything else up to 5 files and 300 changed lines, no risk surface | Pass 1 inline, parent roast, tiered Sin Inventory, approval bar. Codex only if the parent pass ends with nothing confirmed. |\n| **HIGH-RISK** | Any risk surface touched, or more than 5 files or 300 changed lines, or the user asked for a deep review, an exhaustive roast, or a second opinion | The full program: disk-backed evidence plus the Codex cross-model review. |\n\nSet the class from this table before reviewing and announce it before any other\nreview commentary. The class is a program choice, not a severity label. Keep it\nfixed unless the evidence meets the explicit escalation rule below.\n\n**Risk surface outranks size.** A five-line diff that hardcodes a credential,\nloosens an auth check, or builds a query by string concatenation is HIGH-RISK.\nSize can only downgrade a change that touches no risk surface.\n\nAnnounce the class in one line before reviewing, so the user can see which\nprogram ran and override it:\n\n```\nScope: 2 files, 18 lines, no risk surface — SMALL review (single pass, no Codex).\n```\n\n**Escalate on evidence, never on suspicion.** Upgrade mid-review only when the\npass confirms a security, data-loss, or crash-level problem, the diff turns out\nmaterially larger than the stat implied, or the user asks for more depth. Say so\nin one line — `Upgrading to HIGH-RISK: unsanitized input reaches exec at\nsrc\u002Fthumb.ts:44` — and continue with the larger program.\n\nAn ordinary bug that the cheap pass found is the cheap pass working. Fix it,\nreport it, and stay in the class you started in — do not inflate a finding's\nseverity to justify a bigger program.\n\nNever downgrade a depth the user explicitly asked for, and never upgrade because\nthe code looks interesting.\n\n---\n\n## Pass 1: Simplify (auto-fix)\n\nPass 1 runs the vs-deslop on-demand anti-slop file pass on in-scope\nnamed TS\u002FJS files. Follow `..\u002Fvs-deslop\u002FSKILL.md` and run\n`node skills\u002Fvs-deslop\u002Fscripts\u002Frun-anti-slop.mjs` on those named files;\ndo not skip that file pass. Do not report `CLEAN` without that\nnamed-file runner.\n\nClean the code first. The parent performs one integrated reuse, quality, and\nefficiency pass. Delegate separate review domains only when deep effort was\nselected and the shared budget permits it; verify findings before applying fixes.\n\n**SMALL and STANDARD:** read the diff directly (`git diff\n\u003Cresolved-scope-arguments>`) plus the enclosing functions and their callers. Skip\nthe evidence capture below — under 300 changed lines, writing a patch file, an\nindex, and a manifest costs more than the change is worth. Sweep the three lenses\nplus correctness and security inline and fix what is confirmed. SMALL then goes\nstraight to the SMALL Verdict section; STANDARD continues to Pass 2.\n\n**HIGH-RISK:** persist the authoritative diff instead of printing it\ninto model context. Resolve the shared `evidence-manifest.mjs`, write `git diff`\n(or `git diff HEAD` for staged changes) to an evidence file under\n`~\u002F.vs\u002F$PROJECT_ID\u002Freviews\u002Fevidence\u002F`, then emit its manifest and a bounded\n`diff --git` \u002F `@@` hunk index. Inspect only the relevant diff hunks, owning\nsource functions, callers, and tests. This keeps the full review surface\navailable without paying for it on every pass. Follow the\n[disk-backed evidence contract](..\u002Fvs-internal-shared\u002Freferences\u002Fdisk-backed-evidence.md).\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nREVIEW_EVIDENCE_DIR=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Freviews\u002Fevidence\"\nEVIDENCE_TOOL=\"\u003Cresolved-vs-internal-shared-skill-directory>\u002Fscripts\u002Fevidence-manifest.mjs\"\nmkdir -p \"$REVIEW_EVIDENCE_DIR\"\n\nDIFF_PATH=\"$REVIEW_EVIDENCE_DIR\u002Fdiff-$(date +%Y%m%d-%H%M%S).patch\"\nHUNK_INDEX_PATH=\"$DIFF_PATH.index.txt\"\ngit diff --no-color \u003Cresolved-scope-arguments> > \"$DIFF_PATH\"\nrg -n '^(diff --git|@@)' \"$DIFF_PATH\" > \"$HUNK_INDEX_PATH\"\nnode \"$EVIDENCE_TOOL\" manifest \"$DIFF_PATH\" \"$HUNK_INDEX_PATH\"\nhead -200 \"$HUNK_INDEX_PATH\"\n```\n\nReplace `\u003Cresolved-scope-arguments>` from Phase 0 (`--cached`, `main...HEAD`, or\nthe explicit user scope). Search the complete disk-backed hunk index by filename\nor symbol when a relevant hunk falls beyond the initial orientation sample.\n\n### Lens 1: Code Reuse\n\n1. Search for existing utilities that could replace new code\n2. Flag functions duplicating existing functionality — suggest the existing one\n3. Flag inline logic where an existing utility applies\n\n### Lens 2: Code Quality\n\n1. Redundant state \u002F derived values that should be computed\n2. Parameter sprawl instead of restructuring\n3. Copy-paste with variation that should be unified\n4. Leaky abstractions breaking encapsulation\n5. Stringly-typed code where constants\u002Fenums exist\n6. Unnecessary JSX nesting adding no layout value\n7. Comments explaining WHAT (delete; keep only non-obvious WHY)\n8. AI slop: hallucinated imports, verbose boilerplate, defensive nulls on non-null types, wrappers adding zero logic\n\n### Lens 3: Efficiency\n\n1. Redundant computations, repeated reads, duplicate API calls, N+1\n2. Independent operations that could run in parallel\n3. Blocking work on hot paths (startup, per-request, per-render)\n4. No-op state updates in loops\u002Fintervals — add change-detection guards\n5. TOCTOU existence checks — operate directly, handle error\n6. Unbounded data structures, missing cleanup, listener leaks\n7. Reading entire files when only a portion is needed\n\n### Auto-apply\n\nAggregate findings from all three lenses. Fix each issue directly. Skip false positives — don't argue, just move on. Briefly summarize what was fixed.\n\n**Pre-flight:** do not auto-apply. Sweep the lenses and report. Leave the\nworking tree untouched until the user says fix.\n\n**Sweep for the other copies.** Before calling a fix done, grep the repository\nfor further copies of what you just changed — the same CSS selector, constant,\nthreshold, validation rule, or copy-pasted block. A rule that exists twice and is\nfixed once ships the bug in the copy you did not read, and the diff will not show\nit to you. Fix every copy in scope, or name the unfixed ones as a finding.\n\n---\n\n## Pass 2: Parent Roast + Gated Codex Review\n\n**STANDARD and HIGH-RISK only.** A SMALL change already got correctness and\nsecurity coverage in its single pass; a second sweep over 18 lines finds nothing\nthe first one missed.\n\nRun on the cleaned code. The parent owns the roast while Codex provides the\ncross-model second opinion. Do not spawn another local child merely to restate\nthe parent's review.\n\n### Parent Roast\n\nGather intel first: imports\u002Fexports, callers, tests. Then sweep all scan lenses — find what Pass 1 missed:\n- **Correctness** — runtime breakage, logic errors, null access, off-by-one\n- **Security** — injection, unsafe input, missing auth\n- **Architecture** — structural regressions, god objects, circular deps, mixed concerns\n- **Error handling** — swallowed exceptions, silent failures, empty catches\n\nFor the **Architecture** lens, when changed code suggests a structural\nregression, load and follow\n[`..\u002Fvs-architect\u002FSKILL.md`](..\u002Fvs-architect\u002FSKILL.md) in diff-scoped composed\nmode. Use it to deepen confirmed structural findings; do not open its standalone\ncandidate-selection gate, broaden beyond the review scope, or design an\ninterface. Roast Code still owns severity, fixes, and the verdict.\n\nFor every meaningful change, ask whether there is a structural simplification\nthat would make the implementation smaller, more direct, and easier to explain.\nLook for ways to delete whole branches, helpers, modes, wrappers, layers, or\nincidental concepts instead of merely moving them around.\n\n### Ponytail pass\n\nRun this bounded pass in every review class:\n\n1. Check once whether equivalent behavior or a canonical repository helper\n   already exists.\n2. Check whether standard-library or native-platform behavior can replace a\n   new dependency or custom implementation.\n3. Look for an entire dependency, file, wrapper, mode, branch, abstraction, or\n   configuration entry that can disappear while the requested behavior stays\n   complete.\n4. Reject speculative extensibility that serves no explicit requirement.\n\nA Ponytail finding must name the removable machinery, the smaller verified\nalternative, and the behavior-preservation check. Do not report a vague\n\"simplify this\" preference or invent a finding when the chosen machinery is\nalready the first complete rung.\n\nArchitecture findings must be framed as deepening opportunities, not vague \"extract a service\" advice. Include: `Files\u002Fmodules`, `Problem`, `Suggested deepening`, `Test surface`, and `Why this improves locality\u002Fleverage`. If the suggestion does not reduce caller pain or clarify a public seam, do not report it as architecture.\n\nStructural review guidance lives in [structural-review.md](references\u002Fstructural-review.md). Use it for maintainability-heavy reviews or when the diff technically works but feels tangled.\n\nRate each finding with confidence (0-100). Only report 80+.\n\nDeliver 1-3 opening zingers based on worst patterns — one per real finding, never\nmore than the findings support. Reference actual names, line counts. See [comedy-techniques.md](references\u002Fcomedy-techniques.md).\n\n### Codex Review\n\nRun the advisor lane when either trigger fires:\n\n- **HIGH-RISK** — the Phase 1 class already encodes it: an explicit\n  second-opinion request, a diff past 5 files or 300 changed lines, or a touched\n  risk surface.\n- **A clean parent pass on a STANDARD change** — the parent roast confirmed\n  nothing. A clean self-review is the least trustworthy verdict this skill\n  produces, and it is the cheapest moment to buy an independent one: the lane\n  fires only on diffs where there is nothing else left to do.\n\nSkip it for SMALL unless the user asked. Never replace it with a mental review or\na slash-command reference — the point is that a different model read the code.\n\n**In Claude Code:** run the codex plugin's review command:\n```bash\nnode \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcodex-companion.mjs\" review --wait\n```\nIf `CLAUDE_PLUGIN_ROOT` is not set or the script doesn't exist, fall back to\nthe direct CLI below.\n\n**In Codex:** run the shell command directly.\n\n**Derive the advisor scope from the Phase 0 selection — never from global git\nstate and never by probing.** `--uncommitted` reviews the dirty working tree,\nwhile `--base` reviews the branch diff. Preserve the chosen review surface even\nwhen unrelated dirty files exist:\n\n```bash\ncase \"$REVIEW_SCOPE_KIND\" in\n  explicit-uncommitted|staged) CODEX_SCOPE_ARGS=(--uncommitted) ;;\n  explicit-branch|branch) CODEX_SCOPE_ARGS=(--base \"$BASE_BRANCH\") ;;\n  explicit-file) CODEX_SCOPE_ARGS=() ;;\nesac\nCODEX_SCOPE_PROMPT=\"Review only the Phase 0 scope: $REVIEW_SCOPE_DESCRIPTION\"\n\nif [ \"$REVIEW_CLASS\" = \"HIGH-RISK\" ]; then\n  timeout 120 codex review \"${CODEX_SCOPE_ARGS[@]}\" \"$CODEX_SCOPE_PROMPT\" 2>\u002Fdev\u002Fnull \\\n    | node \"$EVIDENCE_TOOL\" capture \"$REVIEW_EVIDENCE_DIR\u002Fcodex-review.txt\"\nelse\n  timeout 120 codex review \"${CODEX_SCOPE_ARGS[@]}\" \"$CODEX_SCOPE_PROMPT\" 2>\u002Fdev\u002Fnull\nfi\n```\n\nRe-resolve the scope only after you changed its source yourself (for example,\nyou committed the staged change mid-review). Running `codex review --help`,\nretrying another flag on empty output, or polling a backgrounded review with\n`sleep` is wasted budget.\n\nFor HIGH-RISK, resolve `$EVIDENCE_TOOL` and `$REVIEW_EVIDENCE_DIR` as in Pass 1\nand inspect the artifact with bounded `slice` ranges around concrete findings.\nFor STANDARD, the command reads advisor stdout directly and does not reference\nthe HIGH-RISK-only evidence variables.\n\n`timeout 120` is the whole retry policy. If `timeout` is unavailable, run the\nreview in the foreground and abort at two minutes — never hand it to a background\ntask and poll. When the executable is missing, the environment blocks it, the\nuser declines to share the diff externally, or it produces nothing within the\nwindow: log the lane as skipped with the reason and honor the missing-lane rule\nbelow.\n\n**Missing-lane rule.** The advisor lane failing does not make the review\nstronger, so the verdict may not read as though it ran:\n\n- Report the lane explicitly in the closeout — `Lanes: codex ✗ (timeout)`.\n- On HIGH-RISK, the verdict is \"reviewed, not independently verified — human\n  review required\". Never `Remaining: 0` on its own.\n- Cap confidence at 70%, exactly as the Zero-Finding Gate does.\n\nThe one confirmed miss this skill has on record happened exactly this way: a\nparent pass closed a review with `Remaining: 0` and no advisor lane, on a diff an\nindependent reviewer then broke into two capital-level and three felony-level\nfindings.\n\n**Re-review only when it earned a re-run.** After fixes, re-run the advisor only\nif a CAPITAL OFFENSE or FELONY was fixed. For lower tiers the focused tests plus\na re-read of the changed hunks are the proof; a second full advisor pass to\nconfirm a misdemeanor is budget spent on wording.\n\nParse Codex review output for finding titles, bodies, priorities, and locations when present. Map Codex priorities to roast severity: P0\u002FP1 = Critical, P2 = Serious, P3 = Minor. If Codex returns unstructured text, summarize the findings manually and note that structured output was unavailable.\n\n---\n\n## SMALL Verdict\n\nA SMALL review answers in a few lines: no tier headers, no worst-offender\nspotlight, no fix menu.\n\nStart with the Phase 1 scope announcement containing the literal words `SMALL\nreview`. Then report only the flat finding list below. Do not print `CAPITAL\nOFFENSES`, `FELONIES`, `CRIMES`, `MISDEMEANORS`, or `PARKING TICKETS` anywhere\nin a SMALL response; those labels mean the wrong program ran.\n\n- One zinger, and only if the code earned it.\n- A flat list of at most 3 findings, worst first:\n  `**Act** — **[Sin Name]** — file:line` (or Consider \u002F Noted \u002F Dismissed)\n  plus one line on what to do. Every finding still carries a bucket; skip empty\n  Act\u002FConsider headers when there are no findings.\n- Fix the confirmed ones directly, then one line on what changed.\n  **Pre-flight:** do not fix. Report the bucketed list and go to closeout.\n- Nothing wrong? Say so, name the one thing that is specifically right with\n  `file:line` evidence, and stop. Do not open a tier to hold a single nitpick.\n\nEscalate to the Sin Inventory below only when the pass confirms one of:\n\n- a security, data-loss, or crash-level problem\n- a provable no-op — the fix cannot execute or affect the path it targets, the\n  guard is checked where it is never armed, or the test asserts something other\n  than the behavior in question\n\nA plain correctness bug, fixed inline with a one-line finding, is a complete\nSMALL review — do not reach for a severity tier to dress it up. A no-op fix is\ndifferent in kind: green tests and a confident summary are actively laundering it,\nso it earns the spotlight even on a twenty-line diff.\n\nWrong output from a path that does execute is an ordinary correctness bug, not a\nno-op and not an escalation trigger.\n\n---\n\n## Sin Inventory\n\n**STANDARD and HIGH-RISK.** Aggregate findings from Roast + Codex. Deduplicate — if both flag the same line, keep the more specific finding. Tag Codex-only findings so the user sees the cross-model signal.\n\nEntry guard: if the Phase 1 class is SMALL, stop and return the SMALL Verdict.\nDo not open this section or use any taxonomy tier label.\n\nGroup by the fixed 5-tier taxonomy — **CAPITAL OFFENSES \u002F FELONIES \u002F CRIMES \u002F MISDEMEANORS \u002F PARKING TICKETS**. Use these exact labels every time; keep the openers and metaphors fresh, keep the tiers stable. Each sin: `N. **[Sin Name]** — file:line` + bucket + one-liner roast; assign tier by impact. The bucket is actionability, not a sixth tier.\n\nDrop a tier that has nothing in it rather than filling it. A finding that is moot\nby its own admission — a guard that can never fire, a nitpick on a line you just\ndeleted — is not a PARKING TICKET, it is padding.\n\nIf 15+ sins, show top 10 by severity. Mention overflow count.\n\nFor tier criteria and the sin-to-tier mapping, see [sin-categories.md](references\u002Fsin-categories.md). Any guidance that contradicts the fixed taxonomy (e.g., \"invent your own labels\") is superseded by this section.\n\nWhen structural problems are present, prioritize them above cosmetic cleanup:\nstructural regressions, missed simplification, spaghetti branching, unclear\ntype\u002Fboundary contracts, file-size\u002Fdecomposition pressure, then smaller\nlegibility issues. Do not bury a design regression under naming or formatting\nnoise.\n\n**Worst offender spotlight:** deep dive on the biggest sin — what it does, what it should be, blast radius.\n\n**Fix CAPITAL OFFENSES + FELONIES automatically and proceed** when they are\nAct. Do not print a tier menu — say which tiers you fixed in one line; the user\ncan interrupt for more or less.\n\n**Pre-flight:** do not auto-fix CAPITAL \u002F FELONY Acts. Report them and go to\ncloseout.\n\n## Review Buckets\n\nEvery finding is bucketed. A roast that lists findings with no buckets fails.\nA roast that puts a nit in Act as if it were blocking fails. Empty Act\u002FConsider\nheaders with no findings are slogan-only and fail.\n\nBuckets are actionability, not severity. Keep the sin tiers. Tag each finding\nwith exactly one bucket:\n\n- **Act** — must change before merge; owner is the PR author \u002F executor\n- **Consider** — judgment call; do not block merge\n- **Noted** — recorded, no action\n- **Dismissed** — seen and rejected, with one-line why\n\nPARKING TICKETS, preference, and formatting nits are never Act. Confirmed\ncorrectness, security, or crash-level sins that would block a real PR are Act.\n\nSMALL stays flat: prefix each of the at most 3 findings with the bucket name.\nSTANDARD and HIGH-RISK keep the Sin Inventory; each sin still carries its\nbucket.\n\nAdvisors still run as today (Codex, compose, the shared independent-advisors\nlane). They do not invent a new channel. The parent assigns the bucket after\nverifying the finding.\n\nScore the roast with `skills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-buckets.mjs`\n(exit 1 is a fail). Exclusive roast cases live under `test\u002Ffixtures\u002Fbuckets`.\n\n## Fix\n\nProcess selected fixes. Show before\u002Fafter for major changes. Run linter if available.\n\n**Pre-flight:** skip this section until the user says fix. Verdict first.\n\nFor each review finding, act from its bucket:\n\n- Act — confirmed in code, fixed or reported; the PR author \u002F\n  executor owns the change\n- Dismissed — not real, too speculative, intentionally designed, or\n  worse than the code it would replace, plus the one-line why\n- Consider or Noted — judgment call or recorded, no merge block\n\nIf a review-triggered fix changes code, rerun the focused tests\u002Fchecks that cover\nthe touched behavior and rerun the relevant review pass. Keep going until there\nare no Act findings left in scope. Once a rerun is clean, stop;\ndo not spend another review cycle just to get nicer closeout wording.\n\nCloseout for STANDARD and HIGH-RISK is one line, not a scorecard block:\n\n```\nPass 1: [N] fixed | Pass 2: [N] found, [M] absolved | Files: N | Lanes: codex ✓ | Remaining: [N] Act\n```\n\n`Lanes` is mandatory and names every evidence lane the class called for, with a\nreason when one is missing: `Lanes: codex ✗ (declined — private diff)`. A verdict\nthat hides which lanes ran is the defect the missing-lane rule exists to prevent.\n\nSMALL closes out in one line: `SMALL review: [N] fixed, [N] left — \u003Cfiles>`.\n\n## Zero-Finding Gate\n\nApplies in every class. If all applicable review passes produce zero findings —\nno Act findings across every lens:\n\n1. Verify you read the changed files in full, not just diffstat.\n2. Name at least one specific positive assertion with `file:line` evidence:\n   \"auth is correct because X at `src\u002Fauth.ts:42`\"\n3. For STANDARD and HIGH-RISK, spend the advisor lane before declaring clean —\n   this is the trigger described under Codex Review. A clean parent pass is a\n   reason to get a second opinion, not a reason to skip one.\n4. If still zero findings after the positive-assertion pass and the advisor lane,\n   cap confidence at 70% and note \"Zero findings — low-confidence approval\" in the\n   summary.\n\nA clean review that can't name what's specifically right is a rubber stamp. A\nclean review that invents a finding rather than say \"clean\" is worse.\n\n## Approval Bar\n\nDo not approve a change merely because behavior appears correct. Passing tests\nare necessary, not a waiver for maintainability debt.\n\nWalk this list for STANDARD and HIGH-RISK. For SMALL, only report a blocker you\nactually tripped over while reading the diff — a 20-line change rarely has a\ndecomposition problem, and hunting for one is how a small review turns into a\nlong one.\n\nTreat these as high-confidence blockers when they are in scope, supported by\nevidence, and materially worsen maintainability:\n\n- the implementation preserves incidental complexity when a clear restructure\n  could remove branches, modes, wrappers, helpers, or layers\n- the diff pushes a file from below 1000 lines to above 1000 lines without a\n  compelling structural reason\n- ad-hoc conditionals or one-off flags are bolted into an already busy flow\n- feature-specific checks get scattered through shared\u002Fgeneral-purpose code\n- an abstraction, wrapper, cast, or optional boundary makes the design more\n  indirect without buying clarity\n- the code duplicates an existing canonical helper or puts logic in a layer\n  that does not own the concept\n- independent async work is serialized, or related updates can leave state\n  half-applied, when a simpler atomic\u002Fparallel structure is obvious\n\nIf a blocker is present, leave explicit actionable feedback and push for a\ncleaner decomposition. If none are present, say what specific structure made the\nchange acceptable.\n\n## Edge Cases\n\n**Good code:** \"I came here to roast, but... would merge without passive-aggressive comments.\"\nCite the specific positive assertions that earned the clean bill.\n\n**Beyond saving:** \"This isn't technical debt, it's technical bankruptcy.\" Shift to triage plan.\n\n**Inherited code:** \"The original author is long gone. You're not on trial — the code is.\"\n\n## References\n\n- [references\u002Fsin-categories.md](references\u002Fsin-categories.md)\n- [references\u002Fcomedy-techniques.md](references\u002Fcomedy-techniques.md)\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: emit **Next** only. Composed: return to caller.\n\n**Prev:** `\u002Fvs-build-it` | `\u002Fvs-tdd` | `\u002Fvs-qa`\n**Next:** `\u002Fvs-ship-it`\n**Relevant:** `\u002Fvs-pushback`\n","# Roast Code\n\nClassify the change, then review it at the depth it deserves. First pass cleans.\nSecond pass roasts what's left.\n\n## Building Block Composition\n\nRoast Code is a building-block review tool. It consumes:\n\n- `vs-deslop` semantics during Pass 1: simplify working code while preserving behavior. Follow `..\u002Fvs-deslop\u002FSKILL.md` and run `node skills\u002Fvs-deslop\u002Fscripts\u002Frun-anti-slop.mjs` on in-scope named TS\u002FJS files; do not skip that file pass. Do not report `CLEAN` without that named-file runner.\n- [`vs-ponytail`](..\u002Fvs-ponytail\u002FSKILL.md) in composed mode when judging\n  whether new machinery is necessary.\n- [independent-advisors](..\u002Fvs-internal-shared\u002Freferences\u002Findependent-advisors.md)\n  during Pass 2: Codex or another model is an independent signal, not the final\n  judge.\n- `vs-verify` after fixes when the cleanup or review change could affect behavior.\n\nBefore delegating, load and follow\n[`..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md`](..\u002Fvs-internal-shared\u002Freferences\u002Fsubagents.md).\n\n## Critical Rules\n\n1. **Respect chat context** — review ONLY files in scope. Never expand uninvited.\n2. **Scale to the change** — classify in Phase 1 and run only that class's\n   program. A two-line fix does not get an evidence file, five tier headers, and\n   a cross-model review.\n3. **Keep SMALL output flat** — after classifying a change as SMALL, announce\n   `SMALL review`, skip Pass 2 and the Sin Inventory, and use no taxonomy tier\n   names anywhere in the response. Finding a normal bug does not unlock tiered\n   ceremony; only a Phase 1 escalation changes the class.\n4. **Never manufacture findings to fill ceremony** — an empty tier, an unused\n   worst-offender spotlight, or an approval bar with no blocker is the correct\n   output when the change is clean. Inventing a structural finding, a severity\n   tier, or a blocker to justify the program you ran is a review defect.\n5. **Own roast requests** — if the user asks to \"roast\" code or changes, use this skill instead of `simplify`.\n6. **Verify before roasting** — only flag what you've confirmed. Being wrong kills comedy.\n7. **Security first — redact, never quote** — secrets, keys, credentials get escalated to the top, before anything else. But when reporting them: cite `file:line` and the variable name, and NEVER output the actual secret value in your response. Quote the shape only, with the value masked (e.g., `API_KEY = \"sk-live-****\"`). The goal is to flag the sin, not to leak the secret into the transcript, the roast, or any follow-up fix plan. This overrides any roast convention about quoting actual code for specificity — for secret values, redact. The ban covers the whole response, not just the finding that flags the secret: masking it once does not license quoting it in a zinger, an aside, or a finding about some other sin. A secret is funniest masked. When committed credentials are present, explicitly tell the user to rotate\u002Frevoke them and move them to env vars or a secret manager.\n8. **Punch up not down** — mock patterns, not people.\n9. **Be specific** — cite `file:line`, quote actual code. Generic roasts are lazy.\n10. **Treat review output as advisory** — verify every Act finding against\n   the real code path and adjacent files before fixing or reporting it as true.\n11. **Reject speculative review noise** — skip unrealistic edge cases, vague\n    rewrites, and fixes that over-complicate the codebase.\n12. **Bucket every finding** — Act, Consider, Noted, or Dismissed. An unbucketed\n    list fails. A nit in Act fails. Advisors stay the shared Codex \u002F compose \u002F\n    independent-advisors lane; they do not invent a new channel.\n13. **Pre-flight is read-only go\u002Fno-go** — when the user asks for a pre-flight,\n    a ready-to-merge check, reasons not to ship, or names a healthy point, pin\n    that point and review. Do not publish. Do not auto-fix until they say fix.\n\n**Tone:** Senior dev who's seen too much + Gordon Ramsay energy. Not mean, not personal. \"I'm roasting because I care.\"\n\n## Pre-flight mode\n\nEnter this mode when the user asks for a pre-flight, a ready-to-merge check,\nreasons not to ship, or names a healthy point (a commit, branch, tag, `HEAD~n`,\nor any moment the app worked). Those phrases stay in this skill. They are not a\npublish and they do not open a new slash skill.\n\n### Pin, then reuse one command\n\nDefault the healthy point to `main` when the user does not name one. Pin it\nbefore any review:\n\n```bash\ngit rev-parse --verify \u003Chealthy-point>\ngit diff \u003Chealthy-point>...HEAD\n```\n\nReuse that same three-dot command as the Phase 0 scope. Capture the resolved\nSHA. Fail-closed in the pin and stop before review when `git rev-parse` cannot\nresolve the healthy point, or when the three-dot diff is empty.\n\nDo not use the ordinary empty-scope prompt (\"forgot to stage\") in this mode. A\nbad or empty pin is not a roast; it is a stop.\n\n### Review as roast, verdict as go\u002Fno-go\n\nKeep the existing lenses, Act \u002F Consider \u002F Noted \u002F Dismissed buckets, and the\nHIGH-RISK advisor lane. Do not import a second Breaking \u002F Incomplete \u002F Conflict\n\u002F Safe taxonomy. A change that would have been Breaking, Incomplete, or Conflict\nis Act or Consider; Safe is Noted.\n\nThis mode is read-only until the user explicitly says fix. Skip Pass 1\nauto-apply, SMALL inline fixes, and Act auto-fix of CAPITAL \u002F FELONY. Sweep the\nlenses and report; leave the working tree untouched.\n\n### Bounded interview (optional)\n\nAfter the review, ask only about unresolved implicit decisions the code cannot\nanswer. Compose the questioning rules from [`..\u002Fvs-pushback\u002FSKILL.md`](..\u002Fvs-pushback\u002FSKILL.md)\n— one question at a time, recommended answer first, Finding \u002F Recommendation \u002F\nImpact. Do not reintroduce `\u002Fgrill-me`. Skip the interview when nothing implicit\nremains.\n\n### Closeout\n\nClose with exactly one verdict:\n\n- **SHIP** — zero Act findings, the healthy point stayed pinned, and the\n  interview is done or was skipped.\n- **DO NOT SHIP** — list every Act with file path, line number, and snippet.\n\nSHIP is illegal while any Act remains, the healthy point is unresolved (missing\npin, bad ref, or empty three-dot diff), or an interview question is still open.\n\nOn SHIP, the standalone handoff is `\u002Fvs-ship-it`. Do not publish from this\nskill. Do not compose this review into `vs-ship-it`.\n\nScore a pre-flight closeout with\n`skills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-preflight.mjs` (exit 1 is a fail).\nExclusive pre-flight cases live under `test\u002Ffixtures\u002Fpreflight`.\n\n## Phase 0: Scope\n\n**Priority:**\n1. Chat context (specific files under discussion)\n2. User-specified files\u002Fdirs\n3. Staged: `git diff --cached --name-only`\n4. Branch diff: `git diff main...HEAD --name-only`\n5. If none: ask\n\nIn pre-flight mode the scope is the pinned three-dot command\n`git diff \u003Chealthy-point>...HEAD`. That replaces the staged \u002F branch \u002F ask\nfallbacks. The pin still runs first and still fail-closes on a bad ref or an\nempty diff.\n\nResolve the scope once and retain its exact diff arguments, base branch, paths,\nand kind for every later phase. Use these scope kinds:\n\n- `explicit-uncommitted` when a chat- or user-selected path is staged,\n  unstaged, or untracked\n- `explicit-branch` when selected paths belong to the committed branch diff\n- `explicit-file` for a current-file review with no diff\n- `staged` for the staged fallback\n- `branch` for the branch-diff fallback\n\nUnrelated working-tree changes never replace the selected scope. If the user\nselected an untracked file, retain it as `explicit-uncommitted`; do not infer\nscope from a tracked-files-only status query later.\n\n**If empty:** \"Nothing to roast. Either your code is perfect (unlikely) or you forgot to stage.\"\n\n**Pre-flight empty or bad pin:** stop before review. Do not ask whether they\nforgot to stage. The pin already failed.\n\n---\n\n## Phase 1: Classify the Change\n\nMeasure before reviewing. One stat read plus a look at the changed lines decides\nwhich program runs, and the whole review is scoped to that program.\n\n```bash\ngit diff --stat \u003Cresolved-scope-arguments>\n```\n\nFor explicitly named files with no diff to measure (a file review rather than a\nchange review), use the files themselves as the size signal.\n\n**Signal 1 — size:** changed files, changed lines, and whether the change is\ndocs, comments, tests, or formatting only.\n\n**Signal 2 — risk surface:** does the change touch auth or authorization,\nsecrets and credentials, crypto, query construction, shell or dynamic execution,\nuntrusted input parsing, persistence or migrations, concurrency, payments, a\npublic API or wire contract, or CI, release, permission, and ownership config?\n\nSniff it deterministically on the changed lines, then confirm by reading the hunks\nit hits — a keyword is a pointer, not a verdict:\n\n```bash\ngit diff -U0 \u003Cresolved-scope-arguments> \\\n  | rg -n '^[+-]' \\\n  | rg -v '^[0-9]+:(\\+\\+\\+|---) ' \\\n  | rg -in 'password|secret|token|api[_-]?key|credential|authn|authz|authoriz|permission|jwt|crypto|exec|spawn|eval\\(|innerHTML|SELECT |INSERT |UPDATE |DELETE |migrat|lock|mutex|Promise\\.all|charge|refund|price'\n```\n\nKeep the pattern tight. A sniff that matches every TypeScript diff escalates\nevery review and buys nothing; public-API and wire-contract changes are a\njudgment call on the hunks, not a keyword.\n\nPaths carry the same signal: `.github\u002Fworkflows\u002F`, `migrations\u002F`, `*.sql`,\n`Dockerfile`, lockfiles, `CODEOWNERS`, and dependency manifests are risk surface\nregardless of how few lines changed.\n\n| Class | When | Program |\n|-------|------|---------|\n| **SMALL** | Docs, comments, or formatting only, or at most 2 files and 30 changed lines — and no risk surface | One integrated pass (Pass 1 lenses plus correctness, run inline). No disk-backed capture, no Codex, no tiers. |\n| **STANDARD** | Everything else up to 5 files and 300 changed lines, no risk surface | Pass 1 inline, parent roast, tiered Sin Inventory, approval bar. Codex only if the parent pass ends with nothing confirmed. |\n| **HIGH-RISK** | Any risk surface touched, or more than 5 files or 300 changed lines, or the user asked for a deep review, an exhaustive roast, or a second opinion | The full program: disk-backed evidence plus the Codex cross-model review. |\n\nSet the class from this table before reviewing and announce it before any other\nreview commentary. The class is a program choice, not a severity label. Keep it\nfixed unless the evidence meets the explicit escalation rule below.\n\n**Risk surface outranks size.** A five-line diff that hardcodes a credential,\nloosens an auth check, or builds a query by string concatenation is HIGH-RISK.\nSize can only downgrade a change that touches no risk surface.\n\nAnnounce the class in one line before reviewing, so the user can see which\nprogram ran and override it:\n\n```\nScope: 2 files, 18 lines, no risk surface — SMALL review (single pass, no Codex).\n```\n\n**Escalate on evidence, never on suspicion.** Upgrade mid-review only when the\npass confirms a security, data-loss, or crash-level problem, the diff turns out\nmaterially larger than the stat implied, or the user asks for more depth. Say so\nin one line — `Upgrading to HIGH-RISK: unsanitized input reaches exec at\nsrc\u002Fthumb.ts:44` — and continue with the larger program.\n\nAn ordinary bug that the cheap pass found is the cheap pass working. Fix it,\nreport it, and stay in the class you started in — do not inflate a finding's\nseverity to justify a bigger program.\n\nNever downgrade a depth the user explicitly asked for, and never upgrade because\nthe code looks interesting.\n\n---\n\n## Pass 1: Simplify (auto-fix)\n\nPass 1 runs the vs-deslop on-demand anti-slop file pass on in-scope\nnamed TS\u002FJS files. Follow `..\u002Fvs-deslop\u002FSKILL.md` and run\n`node skills\u002Fvs-deslop\u002Fscripts\u002Frun-anti-slop.mjs` on those named files;\ndo not skip that file pass. Do not report `CLEAN` without that\nnamed-file runner.\n\nClean the code first. The parent performs one integrated reuse, quality, and\nefficiency pass. Delegate separate review domains only when deep effort was\nselected and the shared budget permits it; verify findings before applying fixes.\n\n**SMALL and STANDARD:** read the diff directly (`git diff\n\u003Cresolved-scope-arguments>`) plus the enclosing functions and their callers. Skip\nthe evidence capture below — under 300 changed lines, writing a patch file, an\nindex, and a manifest costs more than the change is worth. Sweep the three lenses\nplus correctness and security inline and fix what is confirmed. SMALL then goes\nstraight to the SMALL Verdict section; STANDARD continues to Pass 2.\n\n**HIGH-RISK:** persist the authoritative diff instead of printing it\ninto model context. Resolve the shared `evidence-manifest.mjs`, write `git diff`\n(or `git diff HEAD` for staged changes) to an evidence file under\n`~\u002F.vs\u002F$PROJECT_ID\u002Freviews\u002Fevidence\u002F`, then emit its manifest and a bounded\n`diff --git` \u002F `@@` hunk index. Inspect only the relevant diff hunks, owning\nsource functions, callers, and tests. This keeps the full review surface\navailable without paying for it on every pass. Follow the\n[disk-backed evidence contract](..\u002Fvs-internal-shared\u002Freferences\u002Fdisk-backed-evidence.md).\n\n```bash\nPROJECT_ID=$(git config --get remote.origin.url 2>\u002Fdev\u002Fnull \\\n  | sed -E 's#\\.git$##; s#.*[:\u002F]([^\u002F]+\u002F[^\u002F]+)$#\\1#; s#\u002F#-#g')\n[ -z \"$PROJECT_ID\" ] && PROJECT_ID=$(basename \"$PWD\")\nREVIEW_EVIDENCE_DIR=\"$HOME\u002F.vs\u002F$PROJECT_ID\u002Freviews\u002Fevidence\"\nEVIDENCE_TOOL=\"\u003Cresolved-vs-internal-shared-skill-directory>\u002Fscripts\u002Fevidence-manifest.mjs\"\nmkdir -p \"$REVIEW_EVIDENCE_DIR\"\n\nDIFF_PATH=\"$REVIEW_EVIDENCE_DIR\u002Fdiff-$(date +%Y%m%d-%H%M%S).patch\"\nHUNK_INDEX_PATH=\"$DIFF_PATH.index.txt\"\ngit diff --no-color \u003Cresolved-scope-arguments> > \"$DIFF_PATH\"\nrg -n '^(diff --git|@@)' \"$DIFF_PATH\" > \"$HUNK_INDEX_PATH\"\nnode \"$EVIDENCE_TOOL\" manifest \"$DIFF_PATH\" \"$HUNK_INDEX_PATH\"\nhead -200 \"$HUNK_INDEX_PATH\"\n```\n\nReplace `\u003Cresolved-scope-arguments>` from Phase 0 (`--cached`, `main...HEAD`, or\nthe explicit user scope). Search the complete disk-backed hunk index by filename\nor symbol when a relevant hunk falls beyond the initial orientation sample.\n\n### Lens 1: Code Reuse\n\n1. Search for existing utilities that could replace new code\n2. Flag functions duplicating existing functionality — suggest the existing one\n3. Flag inline logic where an existing utility applies\n\n### Lens 2: Code Quality\n\n1. Redundant state \u002F derived values that should be computed\n2. Parameter sprawl instead of restructuring\n3. Copy-paste with variation that should be unified\n4. Leaky abstractions breaking encapsulation\n5. Stringly-typed code where constants\u002Fenums exist\n6. Unnecessary JSX nesting adding no layout value\n7. Comments explaining WHAT (delete; keep only non-obvious WHY)\n8. AI slop: hallucinated imports, verbose boilerplate, defensive nulls on non-null types, wrappers adding zero logic\n\n### Lens 3: Efficiency\n\n1. Redundant computations, repeated reads, duplicate API calls, N+1\n2. Independent operations that could run in parallel\n3. Blocking work on hot paths (startup, per-request, per-render)\n4. No-op state updates in loops\u002Fintervals — add change-detection guards\n5. TOCTOU existence checks — operate directly, handle error\n6. Unbounded data structures, missing cleanup, listener leaks\n7. Reading entire files when only a portion is needed\n\n### Auto-apply\n\nAggregate findings from all three lenses. Fix each issue directly. Skip false positives — don't argue, just move on. Briefly summarize what was fixed.\n\n**Pre-flight:** do not auto-apply. Sweep the lenses and report. Leave the\nworking tree untouched until the user says fix.\n\n**Sweep for the other copies.** Before calling a fix done, grep the repository\nfor further copies of what you just changed — the same CSS selector, constant,\nthreshold, validation rule, or copy-pasted block. A rule that exists twice and is\nfixed once ships the bug in the copy you did not read, and the diff will not show\nit to you. Fix every copy in scope, or name the unfixed ones as a finding.\n\n---\n\n## Pass 2: Parent Roast + Gated Codex Review\n\n**STANDARD and HIGH-RISK only.** A SMALL change already got correctness and\nsecurity coverage in its single pass; a second sweep over 18 lines finds nothing\nthe first one missed.\n\nRun on the cleaned code. The parent owns the roast while Codex provides the\ncross-model second opinion. Do not spawn another local child merely to restate\nthe parent's review.\n\n### Parent Roast\n\nGather intel first: imports\u002Fexports, callers, tests. Then sweep all scan lenses — find what Pass 1 missed:\n- **Correctness** — runtime breakage, logic errors, null access, off-by-one\n- **Security** — injection, unsafe input, missing auth\n- **Architecture** — structural regressions, god objects, circular deps, mixed concerns\n- **Error handling** — swallowed exceptions, silent failures, empty catches\n\nFor the **Architecture** lens, when changed code suggests a structural\nregression, load and follow\n[`..\u002Fvs-architect\u002FSKILL.md`](..\u002Fvs-architect\u002FSKILL.md) in diff-scoped composed\nmode. Use it to deepen confirmed structural findings; do not open its standalone\ncandidate-selection gate, broaden beyond the review scope, or design an\ninterface. Roast Code still owns severity, fixes, and the verdict.\n\nFor every meaningful change, ask whether there is a structural simplification\nthat would make the implementation smaller, more direct, and easier to explain.\nLook for ways to delete whole branches, helpers, modes, wrappers, layers, or\nincidental concepts instead of merely moving them around.\n\n### Ponytail pass\n\nRun this bounded pass in every review class:\n\n1. Check once whether equivalent behavior or a canonical repository helper\n   already exists.\n2. Check whether standard-library or native-platform behavior can replace a\n   new dependency or custom implementation.\n3. Look for an entire dependency, file, wrapper, mode, branch, abstraction, or\n   configuration entry that can disappear while the requested behavior stays\n   complete.\n4. Reject speculative extensibility that serves no explicit requirement.\n\nA Ponytail finding must name the removable machinery, the smaller verified\nalternative, and the behavior-preservation check. Do not report a vague\n\"simplify this\" preference or invent a finding when the chosen machinery is\nalready the first complete rung.\n\nArchitecture findings must be framed as deepening opportunities, not vague \"extract a service\" advice. Include: `Files\u002Fmodules`, `Problem`, `Suggested deepening`, `Test surface`, and `Why this improves locality\u002Fleverage`. If the suggestion does not reduce caller pain or clarify a public seam, do not report it as architecture.\n\nStructural review guidance lives in [structural-review.md](references\u002Fstructural-review.md). Use it for maintainability-heavy reviews or when the diff technically works but feels tangled.\n\nRate each finding with confidence (0-100). Only report 80+.\n\nDeliver 1-3 opening zingers based on worst patterns — one per real finding, never\nmore than the findings support. Reference actual names, line counts. See [comedy-techniques.md](references\u002Fcomedy-techniques.md).\n\n### Codex Review\n\nRun the advisor lane when either trigger fires:\n\n- **HIGH-RISK** — the Phase 1 class already encodes it: an explicit\n  second-opinion request, a diff past 5 files or 300 changed lines, or a touched\n  risk surface.\n- **A clean parent pass on a STANDARD change** — the parent roast confirmed\n  nothing. A clean self-review is the least trustworthy verdict this skill\n  produces, and it is the cheapest moment to buy an independent one: the lane\n  fires only on diffs where there is nothing else left to do.\n\nSkip it for SMALL unless the user asked. Never replace it with a mental review or\na slash-command reference — the point is that a different model read the code.\n\n**In Claude Code:** run the codex plugin's review command:\n```bash\nnode \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Fcodex-companion.mjs\" review --wait\n```\nIf `CLAUDE_PLUGIN_ROOT` is not set or the script doesn't exist, fall back to\nthe direct CLI below.\n\n**In Codex:** run the shell command directly.\n\n**Derive the advisor scope from the Phase 0 selection — never from global git\nstate and never by probing.** `--uncommitted` reviews the dirty working tree,\nwhile `--base` reviews the branch diff. Preserve the chosen review surface even\nwhen unrelated dirty files exist:\n\n```bash\ncase \"$REVIEW_SCOPE_KIND\" in\n  explicit-uncommitted|staged) CODEX_SCOPE_ARGS=(--uncommitted) ;;\n  explicit-branch|branch) CODEX_SCOPE_ARGS=(--base \"$BASE_BRANCH\") ;;\n  explicit-file) CODEX_SCOPE_ARGS=() ;;\nesac\nCODEX_SCOPE_PROMPT=\"Review only the Phase 0 scope: $REVIEW_SCOPE_DESCRIPTION\"\n\nif [ \"$REVIEW_CLASS\" = \"HIGH-RISK\" ]; then\n  timeout 120 codex review \"${CODEX_SCOPE_ARGS[@]}\" \"$CODEX_SCOPE_PROMPT\" 2>\u002Fdev\u002Fnull \\\n    | node \"$EVIDENCE_TOOL\" capture \"$REVIEW_EVIDENCE_DIR\u002Fcodex-review.txt\"\nelse\n  timeout 120 codex review \"${CODEX_SCOPE_ARGS[@]}\" \"$CODEX_SCOPE_PROMPT\" 2>\u002Fdev\u002Fnull\nfi\n```\n\nRe-resolve the scope only after you changed its source yourself (for example,\nyou committed the staged change mid-review). Running `codex review --help`,\nretrying another flag on empty output, or polling a backgrounded review with\n`sleep` is wasted budget.\n\nFor HIGH-RISK, resolve `$EVIDENCE_TOOL` and `$REVIEW_EVIDENCE_DIR` as in Pass 1\nand inspect the artifact with bounded `slice` ranges around concrete findings.\nFor STANDARD, the command reads advisor stdout directly and does not reference\nthe HIGH-RISK-only evidence variables.\n\n`timeout 120` is the whole retry policy. If `timeout` is unavailable, run the\nreview in the foreground and abort at two minutes — never hand it to a background\ntask and poll. When the executable is missing, the environment blocks it, the\nuser declines to share the diff externally, or it produces nothing within the\nwindow: log the lane as skipped with the reason and honor the missing-lane rule\nbelow.\n\n**Missing-lane rule.** The advisor lane failing does not make the review\nstronger, so the verdict may not read as though it ran:\n\n- Report the lane explicitly in the closeout — `Lanes: codex ✗ (timeout)`.\n- On HIGH-RISK, the verdict is \"reviewed, not independently verified — human\n  review required\". Never `Remaining: 0` on its own.\n- Cap confidence at 70%, exactly as the Zero-Finding Gate does.\n\nThe one confirmed miss this skill has on record happened exactly this way: a\nparent pass closed a review with `Remaining: 0` and no advisor lane, on a diff an\nindependent reviewer then broke into two capital-level and three felony-level\nfindings.\n\n**Re-review only when it earned a re-run.** After fixes, re-run the advisor only\nif a CAPITAL OFFENSE or FELONY was fixed. For lower tiers the focused tests plus\na re-read of the changed hunks are the proof; a second full advisor pass to\nconfirm a misdemeanor is budget spent on wording.\n\nParse Codex review output for finding titles, bodies, priorities, and locations when present. Map Codex priorities to roast severity: P0\u002FP1 = Critical, P2 = Serious, P3 = Minor. If Codex returns unstructured text, summarize the findings manually and note that structured output was unavailable.\n\n---\n\n## SMALL Verdict\n\nA SMALL review answers in a few lines: no tier headers, no worst-offender\nspotlight, no fix menu.\n\nStart with the Phase 1 scope announcement containing the literal words `SMALL\nreview`. Then report only the flat finding list below. Do not print `CAPITAL\nOFFENSES`, `FELONIES`, `CRIMES`, `MISDEMEANORS`, or `PARKING TICKETS` anywhere\nin a SMALL response; those labels mean the wrong program ran.\n\n- One zinger, and only if the code earned it.\n- A flat list of at most 3 findings, worst first:\n  `**Act** — **[Sin Name]** — file:line` (or Consider \u002F Noted \u002F Dismissed)\n  plus one line on what to do. Every finding still carries a bucket; skip empty\n  Act\u002FConsider headers when there are no findings.\n- Fix the confirmed ones directly, then one line on what changed.\n  **Pre-flight:** do not fix. Report the bucketed list and go to closeout.\n- Nothing wrong? Say so, name the one thing that is specifically right with\n  `file:line` evidence, and stop. Do not open a tier to hold a single nitpick.\n\nEscalate to the Sin Inventory below only when the pass confirms one of:\n\n- a security, data-loss, or crash-level problem\n- a provable no-op — the fix cannot execute or affect the path it targets, the\n  guard is checked where it is never armed, or the test asserts something other\n  than the behavior in question\n\nA plain correctness bug, fixed inline with a one-line finding, is a complete\nSMALL review — do not reach for a severity tier to dress it up. A no-op fix is\ndifferent in kind: green tests and a confident summary are actively laundering it,\nso it earns the spotlight even on a twenty-line diff.\n\nWrong output from a path that does execute is an ordinary correctness bug, not a\nno-op and not an escalation trigger.\n\n---\n\n## Sin Inventory\n\n**STANDARD and HIGH-RISK.** Aggregate findings from Roast + Codex. Deduplicate — if both flag the same line, keep the more specific finding. Tag Codex-only findings so the user sees the cross-model signal.\n\nEntry guard: if the Phase 1 class is SMALL, stop and return the SMALL Verdict.\nDo not open this section or use any taxonomy tier label.\n\nGroup by the fixed 5-tier taxonomy — **CAPITAL OFFENSES \u002F FELONIES \u002F CRIMES \u002F MISDEMEANORS \u002F PARKING TICKETS**. Use these exact labels every time; keep the openers and metaphors fresh, keep the tiers stable. Each sin: `N. **[Sin Name]** — file:line` + bucket + one-liner roast; assign tier by impact. The bucket is actionability, not a sixth tier.\n\nDrop a tier that has nothing in it rather than filling it. A finding that is moot\nby its own admission — a guard that can never fire, a nitpick on a line you just\ndeleted — is not a PARKING TICKET, it is padding.\n\nIf 15+ sins, show top 10 by severity. Mention overflow count.\n\nFor tier criteria and the sin-to-tier mapping, see [sin-categories.md](references\u002Fsin-categories.md). Any guidance that contradicts the fixed taxonomy (e.g., \"invent your own labels\") is superseded by this section.\n\nWhen structural problems are present, prioritize them above cosmetic cleanup:\nstructural regressions, missed simplification, spaghetti branching, unclear\ntype\u002Fboundary contracts, file-size\u002Fdecomposition pressure, then smaller\nlegibility issues. Do not bury a design regression under naming or formatting\nnoise.\n\n**Worst offender spotlight:** deep dive on the biggest sin — what it does, what it should be, blast radius.\n\n**Fix CAPITAL OFFENSES + FELONIES automatically and proceed** when they are\nAct. Do not print a tier menu — say which tiers you fixed in one line; the user\ncan interrupt for more or less.\n\n**Pre-flight:** do not auto-fix CAPITAL \u002F FELONY Acts. Report them and go to\ncloseout.\n\n## Review Buckets\n\nEvery finding is bucketed. A roast that lists findings with no buckets fails.\nA roast that puts a nit in Act as if it were blocking fails. Empty Act\u002FConsider\nheaders with no findings are slogan-only and fail.\n\nBuckets are actionability, not severity. Keep the sin tiers. Tag each finding\nwith exactly one bucket:\n\n- **Act** — must change before merge; owner is the PR author \u002F executor\n- **Consider** — judgment call; do not block merge\n- **Noted** — recorded, no action\n- **Dismissed** — seen and rejected, with one-line why\n\nPARKING TICKETS, preference, and formatting nits are never Act. Confirmed\ncorrectness, security, or crash-level sins that would block a real PR are Act.\n\nSMALL stays flat: prefix each of the at most 3 findings with the bucket name.\nSTANDARD and HIGH-RISK keep the Sin Inventory; each sin still carries its\nbucket.\n\nAdvisors still run as today (Codex, compose, the shared independent-advisors\nlane). They do not invent a new channel. The parent assigns the bucket after\nverifying the finding.\n\nScore the roast with `skills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-buckets.mjs`\n(exit 1 is a fail). Exclusive roast cases live under `test\u002Ffixtures\u002Fbuckets`.\n\n## Fix\n\nProcess selected fixes. Show before\u002Fafter for major changes. Run linter if available.\n\n**Pre-flight:** skip this section until the user says fix. Verdict first.\n\nFor each review finding, act from its bucket:\n\n- Act — confirmed in code, fixed or reported; the PR author \u002F\n  executor owns the change\n- Dismissed — not real, too speculative, intentionally designed, or\n  worse than the code it would replace, plus the one-line why\n- Consider or Noted — judgment call or recorded, no merge block\n\nIf a review-triggered fix changes code, rerun the focused tests\u002Fchecks that cover\nthe touched behavior and rerun the relevant review pass. Keep going until there\nare no Act findings left in scope. Once a rerun is clean, stop;\ndo not spend another review cycle just to get nicer closeout wording.\n\nCloseout for STANDARD and HIGH-RISK is one line, not a scorecard block:\n\n```\nPass 1: [N] fixed | Pass 2: [N] found, [M] absolved | Files: N | Lanes: codex ✓ | Remaining: [N] Act\n```\n\n`Lanes` is mandatory and names every evidence lane the class called for, with a\nreason when one is missing: `Lanes: codex ✗ (declined — private diff)`. A verdict\nthat hides which lanes ran is the defect the missing-lane rule exists to prevent.\n\nSMALL closes out in one line: `SMALL review: [N] fixed, [N] left — \u003Cfiles>`.\n\n## Zero-Finding Gate\n\nApplies in every class. If all applicable review passes produce zero findings —\nno Act findings across every lens:\n\n1. Verify you read the changed files in full, not just diffstat.\n2. Name at least one specific positive assertion with `file:line` evidence:\n   \"auth is correct because X at `src\u002Fauth.ts:42`\"\n3. For STANDARD and HIGH-RISK, spend the advisor lane before declaring clean —\n   this is the trigger described under Codex Review. A clean parent pass is a\n   reason to get a second opinion, not a reason to skip one.\n4. If still zero findings after the positive-assertion pass and the advisor lane,\n   cap confidence at 70% and note \"Zero findings — low-confidence approval\" in the\n   summary.\n\nA clean review that can't name what's specifically right is a rubber stamp. A\nclean review that invents a finding rather than say \"clean\" is worse.\n\n## Approval Bar\n\nDo not approve a change merely because behavior appears correct. Passing tests\nare necessary, not a waiver for maintainability debt.\n\nWalk this list for STANDARD and HIGH-RISK. For SMALL, only report a blocker you\nactually tripped over while reading the diff — a 20-line change rarely has a\ndecomposition problem, and hunting for one is how a small review turns into a\nlong one.\n\nTreat these as high-confidence blockers when they are in scope, supported by\nevidence, and materially worsen maintainability:\n\n- the implementation preserves incidental complexity when a clear restructure\n  could remove branches, modes, wrappers, helpers, or layers\n- the diff pushes a file from below 1000 lines to above 1000 lines without a\n  compelling structural reason\n- ad-hoc conditionals or one-off flags are bolted into an already busy flow\n- feature-specific checks get scattered through shared\u002Fgeneral-purpose code\n- an abstraction, wrapper, cast, or optional boundary makes the design more\n  indirect without buying clarity\n- the code duplicates an existing canonical helper or puts logic in a layer\n  that does not own the concept\n- independent async work is serialized, or related updates can leave state\n  half-applied, when a simpler atomic\u002Fparallel structure is obvious\n\nIf a blocker is present, leave explicit actionable feedback and push for a\ncleaner decomposition. If none are present, say what specific structure made the\nchange acceptable.\n\n## Edge Cases\n\n**Good code:** \"I came here to roast, but... would merge without passive-aggressive comments.\"\nCite the specific positive assertions that earned the clean bill.\n\n**Beyond saving:** \"This isn't technical debt, it's technical bankruptcy.\" Shift to triage plan.\n\n**Inherited code:** \"The original author is long gone. You're not on trial — the code is.\"\n\n## References\n\n- [references\u002Fsin-categories.md](references\u002Fsin-categories.md)\n- [references\u002Fcomedy-techniques.md](references\u002Fcomedy-techniques.md)\n\n## Output style\n\nApply the [shared output style](..\u002Fvs-internal-shared\u002Freferences\u002Foutput-style.md)\nto every user-facing message.\n\n## Workflow\n\nDirect: emit **Next** only. Composed: return to caller.\n\n**Prev:** `\u002Fvs-build-it` | `\u002Fvs-tdd` | `\u002Fvs-qa`\n**Next:** `\u002Fvs-ship-it`\n**Relevant:** `\u002Fvs-pushback`\n",{"name":5,"description":7},"master","72448991571c9e5f0e76cccf189b8522d442e6eb","skills\u002Fvs-roast-code",1,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs","https:\u002F\u002Fgithub.com\u002Fvltansky","2026-11-06 18:47:54",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add vltansky\u002Fvs --skill vs-roast-code",{"tool":37,"command":38},"bunx","bunx skills add vltansky\u002Fvs --skill vs-roast-code",{"tool":40,"command":41},"pnpm","pnpm dlx skills add vltansky\u002Fvs --skill vs-roast-code","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002FSKILL.md",true,[45,51,55,59,63,69,73,77,81,85,89,93,97,101,105,109,113,117,121,125,129,133,137,141,145,149,153,157,161,165,169,173,177,181,185,189,193,197,201,205,209,213,217,221,225,229,233,237,241,245,249,253,257,261],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"agents\u002Fopenai.yaml","REFERENCE","参考文档",160,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Fagents\u002Fopenai.yaml",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"references\u002Fcomedy-techniques.md",3680,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Freferences\u002Fcomedy-techniques.md",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Fsin-categories.md",3589,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Freferences\u002Fsin-categories.md",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Fstructural-review.md",3433,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Freferences\u002Fstructural-review.md",{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"scripts\u002Freject-roast-buckets.mjs","SCRIPT","脚本",5481,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-buckets.mjs",{"path":70,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":71,"sourceHtmlUrl":72},"scripts\u002Freject-roast-preflight.mjs",9271,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Fscripts\u002Freject-roast-preflight.mjs",{"path":74,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":75,"sourceHtmlUrl":76},"test\u002Fbuckets.static.eval.ts",3795,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Fbuckets.static.eval.ts",{"path":78,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":79,"sourceHtmlUrl":80},"test\u002Fclassification.static.eval.ts",6813,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Fclassification.static.eval.ts",{"path":82,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":83,"sourceHtmlUrl":84},"test\u002Ffixtures\u002Fbuckets\u002Fbad-all-act-nits\u002Froast.md",221,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fbad-all-act-nits\u002Froast.md",{"path":86,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":87,"sourceHtmlUrl":88},"test\u002Ffixtures\u002Fbuckets\u002Fbad-dismissed-as-act\u002Froast.md",271,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fbad-dismissed-as-act\u002Froast.md",{"path":90,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":91,"sourceHtmlUrl":92},"test\u002Ffixtures\u002Fbuckets\u002Fbad-missing-act\u002Froast.md",274,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fbad-missing-act\u002Froast.md",{"path":94,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":95,"sourceHtmlUrl":96},"test\u002Ffixtures\u002Fbuckets\u002Fbad-slogan-only-roast\u002Froast.md",174,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fbad-slogan-only-roast\u002Froast.md",{"path":98,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":99,"sourceHtmlUrl":100},"test\u002Ffixtures\u002Fbuckets\u002Fbad-unbucketed\u002Froast.md",197,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fbad-unbucketed\u002Froast.md",{"path":102,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":103,"sourceHtmlUrl":104},"test\u002Ffixtures\u002Fbuckets\u002Fclean-act-only\u002Froast.md",178,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fclean-act-only\u002Froast.md",{"path":106,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":107,"sourceHtmlUrl":108},"test\u002Ffixtures\u002Fbuckets\u002Fclean-mixed-buckets\u002Froast.md",369,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fclean-mixed-buckets\u002Froast.md",{"path":110,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":111,"sourceHtmlUrl":112},"test\u002Ffixtures\u002Fbuckets\u002Fcopied-slogans-skill.md",163,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fcopied-slogans-skill.md",{"path":114,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":115,"sourceHtmlUrl":116},"test\u002Ffixtures\u002Fbuckets\u002Fphrase-copy-skill.md",751,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fphrase-copy-skill.md",{"path":118,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":119,"sourceHtmlUrl":120},"test\u002Ffixtures\u002Fbuckets\u002Fslogan-only-skill.md",133,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fslogan-only-skill.md",{"path":122,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":123,"sourceHtmlUrl":124},"test\u002Ffixtures\u002Fbuckets\u002Fstructure-paste-skill.md",248,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fbuckets\u002Fstructure-paste-skill.md",{"path":126,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":127,"sourceHtmlUrl":128},"test\u002Ffixtures\u002Fponytail-bloat\u002Fpackage.json",131,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fponytail-bloat\u002Fpackage.json",{"path":130,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":131,"sourceHtmlUrl":132},"test\u002Ffixtures\u002Fponytail-bloat\u002Fsrc\u002Fbloated-name.js",317,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fponytail-bloat\u002Fsrc\u002Fbloated-name.js",{"path":134,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":135,"sourceHtmlUrl":136},"test\u002Ffixtures\u002Fponytail-bloat\u002Fsrc\u002Fdirect-name.js",99,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fponytail-bloat\u002Fsrc\u002Fdirect-name.js",{"path":138,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":139,"sourceHtmlUrl":140},"test\u002Ffixtures\u002Fpreflight\u002Fapply-fix-no-ask\u002Fpreflight.md",319,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fapply-fix-no-ask\u002Fpreflight.md",{"path":142,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":143,"sourceHtmlUrl":144},"test\u002Ffixtures\u002Fpreflight\u002Fauto-fix\u002Fpreflight.md",184,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fauto-fix\u002Fpreflight.md",{"path":146,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":147,"sourceHtmlUrl":148},"test\u002Ffixtures\u002Fpreflight\u002Fbad-ref\u002Fpreflight.md",142,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fbad-ref\u002Fpreflight.md",{"path":150,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":151,"sourceHtmlUrl":152},"test\u002Ffixtures\u002Fpreflight\u002Fclean-ship\u002Fpreflight.md",302,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fclean-ship\u002Fpreflight.md",{"path":154,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":155,"sourceHtmlUrl":156},"test\u002Ffixtures\u002Fpreflight\u002Fcopied-slogans-skill.md",136,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fcopied-slogans-skill.md",{"path":158,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":159,"sourceHtmlUrl":160},"test\u002Ffixtures\u002Fpreflight\u002Fdo-not-ship-act\u002Fpreflight.md",244,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fdo-not-ship-act\u002Fpreflight.md",{"path":162,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":163,"sourceHtmlUrl":164},"test\u002Ffixtures\u002Fpreflight\u002Fempty-diff-waiver\u002Fpreflight.md",199,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fempty-diff-waiver\u002Fpreflight.md",{"path":166,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":167,"sourceHtmlUrl":168},"test\u002Ffixtures\u002Fpreflight\u002Fempty-diff\u002Fpreflight.md",191,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fempty-diff\u002Fpreflight.md",{"path":170,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":171,"sourceHtmlUrl":172},"test\u002Ffixtures\u002Fpreflight\u002Ffix-it-no-ask\u002Fpreflight.md",321,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Ffix-it-no-ask\u002Fpreflight.md",{"path":174,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":175,"sourceHtmlUrl":176},"test\u002Ffixtures\u002Fpreflight\u002Fforged-pin\u002Fpreflight.md",165,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fforged-pin\u002Fpreflight.md",{"path":178,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":179,"sourceHtmlUrl":180},"test\u002Ffixtures\u002Fpreflight\u002Fmutate-no-ask\u002Fpreflight.md",327,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fmutate-no-ask\u002Fpreflight.md",{"path":182,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":183,"sourceHtmlUrl":184},"test\u002Ffixtures\u002Fpreflight\u002Fphrase-copy-skill.md",272,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fphrase-copy-skill.md",{"path":186,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":187,"sourceHtmlUrl":188},"test\u002Ffixtures\u002Fpreflight\u002Fpublish-invoke-skill.md",169,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fpublish-invoke-skill.md",{"path":190,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":191,"sourceHtmlUrl":192},"test\u002Ffixtures\u002Fpreflight\u002Fpublish-run\u002Fpreflight.md",209,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fpublish-run\u002Fpreflight.md",{"path":194,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":195,"sourceHtmlUrl":196},"test\u002Ffixtures\u002Fpreflight\u002Fship-open-interview\u002Fpreflight.md",305,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fship-open-interview\u002Fpreflight.md",{"path":198,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":199,"sourceHtmlUrl":200},"test\u002Ffixtures\u002Fpreflight\u002Fship-unresolved-pin\u002Fpreflight.md",92,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fship-unresolved-pin\u002Fpreflight.md",{"path":202,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":203,"sourceHtmlUrl":204},"test\u002Ffixtures\u002Fpreflight\u002Fship-with-act\u002Fpreflight.md",255,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fship-with-act\u002Fpreflight.md",{"path":206,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":207,"sourceHtmlUrl":208},"test\u002Ffixtures\u002Fpreflight\u002Fslogan-only-skill.md",140,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fslogan-only-skill.md",{"path":210,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":211,"sourceHtmlUrl":212},"test\u002Ffixtures\u002Fpreflight\u002Fstructure-paste-skill.md",252,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fstructure-paste-skill.md",{"path":214,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":215,"sourceHtmlUrl":216},"test\u002Ffixtures\u002Fpreflight\u002Fvs-pre-flight-skill.md",124,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fpreflight\u002Fvs-pre-flight-skill.md",{"path":218,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":219,"sourceHtmlUrl":220},"test\u002Ffixtures\u002Fprioritized-review\u002Fauth.ts",494,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fprioritized-review\u002Fauth.ts",{"path":222,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":223,"sourceHtmlUrl":224},"test\u002Ffixtures\u002Fprioritized-review\u002Fprofile-cache.ts",530,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fprioritized-review\u002Fprofile-cache.ts",{"path":226,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":227,"sourceHtmlUrl":228},"test\u002Ffixtures\u002Fprioritized-review\u002Freport.ts",119,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fprioritized-review\u002Freport.ts",{"path":230,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":231,"sourceHtmlUrl":232},"test\u002Ffixtures\u002Fprioritized-review\u002Fthumbnail.ts",240,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fprioritized-review\u002Fthumbnail.ts",{"path":234,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":235,"sourceHtmlUrl":236},"test\u002Ffixtures\u002Fsecret-code\u002Fconfig.ts",817,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fsecret-code\u002Fconfig.ts",{"path":238,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":239,"sourceHtmlUrl":240},"test\u002Ffixtures\u002Fsinful-code\u002Fdb.ts",70,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fsinful-code\u002Fdb.ts",{"path":242,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":243,"sourceHtmlUrl":244},"test\u002Ffixtures\u002Fsinful-code\u002Fuser-service.ts",1846,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fsinful-code\u002Fuser-service.ts",{"path":246,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":247,"sourceHtmlUrl":248},"test\u002Ffixtures\u002Fsmall-change\u002Fsrc\u002Fduration.js",226,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fsmall-change\u002Fsrc\u002Fduration.js",{"path":250,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":251,"sourceHtmlUrl":252},"test\u002Ffixtures\u002Fsmall-change\u002Fsrc\u002Fduration.test.js",205,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Ffixtures\u002Fsmall-change\u002Fsrc\u002Fduration.test.js",{"path":254,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":255,"sourceHtmlUrl":256},"test\u002Fponytail-pass.eval.ts",2691,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Fponytail-pass.eval.ts",{"path":258,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":259,"sourceHtmlUrl":260},"test\u002Fpreflight.static.eval.ts",6233,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Fpreflight.static.eval.ts",{"path":262,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":263,"sourceHtmlUrl":264},"test\u002Froast-code.eval.ts",20286,"https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-roast-code\u002Ftest\u002Froast-code.eval.ts",[266,273,279,284,291,296],{"skillNo":267,"name":268,"nameZh":6,"description":269,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":270,"categories":271,"keywords":272,"publishedAt":19},"01M77XSKWFMPKFXM0311QKEJHH","vs-write","Use when asked to write, rewrite, edit, tighten, simplify, or polish prose for clarity, or to put text in STE, ASD-STE100, or controlled language. Produces plain, direct, accessible copy without losing substance or precision.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-write\u002FSKILL.md",[],[],{"skillNo":274,"name":275,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":276,"categories":277,"keywords":278,"publishedAt":19},"01M77XSJNX2V560HSKQKQD67KP","vs-fake-verify","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstub-rejector\u002FSKILL.md",[],[],{"skillNo":280,"name":275,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":281,"categories":282,"keywords":283,"publishedAt":19},"01M77XSHHBEMMYZ6B4PD384NZH","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fstructure-paste\u002FSKILL.md",[],[],{"skillNo":285,"name":286,"nameZh":6,"description":287,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":288,"categories":289,"keywords":290,"publishedAt":19},"01M77XSE2Y7MW0KP3DTFNEE2E8","vs-verify","Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS\u002FWARN\u002FFAIL status.","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fpublished-pair\u002FSKILL.md",[],[],{"skillNo":292,"name":275,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":293,"categories":294,"keywords":295,"publishedAt":19},"01M77XSAXFQXNE6VCRC7275Z0X","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002Ftest\u002Ffixtures\u002Fpath-end-state\u002Fphrase-complete-beside\u002FSKILL.md",[],[],{"skillNo":297,"name":286,"nameZh":6,"description":287,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":298,"categories":299,"keywords":300,"publishedAt":19},"01M77XS6R0E9X1E5KKSS00WYSS","https:\u002F\u002Fgithub.com\u002Fvltansky\u002Fvs\u002Fblob\u002Fmaster\u002Fskills\u002Fvs-verify\u002FSKILL.md",[],[]]